Developing a spatial clustering model to identify dangerous intersections in Virginia

摘要

Motor vehicle crashes in Virginia have become a rising issue in the past years. In an effort to reduce crash accidents, this paper aims to identify which intersections in Virginia roadway are the most dangerous, and why they are dangerous. We developed a new distance-based spatial clustering model to group crashes occurred in Virginia roadway into clusters, from which we find those intersections which have the most crashes. In addition, we also gave weights to different types of crashes along with the number of crashes per intersection to define dangerous intersections. Our analysis includes identifying the most dangerous intersections by each year as well as intersections that are specifically more dangerous with regards to alcohol related crashes or speed related crashes. Our findings indicate that the intersections that had high number of crashes were mostly located in the cities such as the Northern Virginia area and Virginia Beach area. After dangerous intersections were identified, we utilized various machine learning methods such as logistic regression, support vector machines and random forest to pinpoint the factors that makes an intersection very dangerous. This information can be beneficial to prevent and reduce crash accidents in the future as well as to raise public awareness.